Skip to content

AbhishekGupta0164/ForgeFlow-Multi-Agentic-AI-System

BranchesTags

Repository files navigation

ForgeFlow Fleet

ForgeFlow Fleet Banner

Not an AI assistant. An AI engineering fleet.

⚡ Parallel Agents • 🧠 AI-Native • 🔧 Developer First • 🚀 Production Ready

📖 Table of Contents

🚀 Overview

ForgeFlow Fleet is a multi-agent AI engineering system that turns messy error logs, screenshots, or vague ideas into complete, production-ready Vercel AI agent blueprints — debugged, designed, and ready to ship — in under 60 seconds.

What It Does

  • 🧩 Debugs production issues from cryptic logs
  • 🏗️ Generates full-stack blueprints for AI agents
  • ⚡ Runs 5 parallel AI specialists simultaneously
  • 🔌 Uses custom MCP servers for intelligent tooling
  • 🖼️ Processes multimodal inputs (text, images, logs, voice)

🎯 The Problem

Developers face two daily productivity killers:

  1. Production Debugging Hell
    Cryptic error logs and stack traces waste hours of investigation time

  2. Idea-to-Implementation Gap
    Turning concepts into structured AI agent architectures is slow and error-prone

ForgeFlow's Solution:
Deploy a 5-specialist AI fleet that works in parallel using the Vercel AI SDK, delivering comprehensive solutions in real-time.

🤖 AI Agent Fleet

Five specialized AI agents working in parallel:

Agent Role Output
🔍 Analyzer Root cause analysis & idea refinement Incident summary, severity assessment, technical context
🎨 UI Architect Interface design & v0 generation 2-3 production-ready v0.dev prompts
⚙️ Backend Engineer System architecture & tooling Architecture blueprint, 3-5 custom MCP tools
⚖️ Judge Solution quality evaluation Usefulness, execution, creativity scores (1-10)
🎬 Scriptwriter Demo creation & storytelling 45-60 second pitch-perfect demo script

✨ Key Features

🚀 Parallel AI Execution

Five specialized agents run simultaneously using Promise.all(), not sequentially. Watch results stream in real-time as each agent completes.

🎨 Built on Modern AI Stack

  • v0.dev: UI components designed with AI-generated prompts
  • Vercel AI SDK: Structured outputs with streamObject and Zod schemas
  • Custom MCP Server: Production-ready tools (log parser, error analyzer, fix generator)

🖼️ Multimodal Intelligence

  • Paste error logs from production
  • Upload screenshots of broken UIs (vision analysis)
  • Type natural language ideas
  • Voice input support (coming soon)

🔄 Real-Time Streaming

Live streaming results powered by Vercel AI SDK's readStreamableValue for instant feedback.

🛠️ Tech Stack

Layer Technology Purpose
Frontend Next.js 15 (App Router) Edge-optimized, streaming-ready dashboard
Styling Tailwind CSS + Framer Motion Responsive design with smooth animations
AI Orchestration Vercel AI SDK 5 parallel agent streams via Server Actions
AI Models Claude 3.5 Sonnet (Anthropic) Deep reasoning and dynamic prompt generation
Persistence Vercel KV Incident history and prompt versioning
Deployment Vercel Global edge deployment
Language TypeScript Type-safe development

🚀 Quick Start

Prerequisites

Installation

# Clone the repository
git clone https://github.com/AbhishekGupta0164/ForgeFlow-Multi-Agentic-AI-System
cd ForgeFlow-Multi-Agentic-AI-System

# Install dependencies
npm install

# Set up environment variables
cp .env.local.example .env.local

Environment Configuration

Edit .env.local and add your API key:

ANTHROPIC_API_KEY=sk-ant-...

Alternative: You can add your API key directly from the UI in Settings → API Keys tab.

Run Development Server

npm run dev

⚙️ How It Works

  1. User Input
    Enter text, logs, screenshots, or voice commands into the Input Card

  2. Command Processing
    Special commands (like /evolve) trigger advanced workflows

  3. Server Action
    launchFleetAction orchestrates 5 parallel streamObject calls

  4. AI Execution
    Each agent generates structured JSON outputs simultaneously

  5. Real-Time Streaming
    Client consumes streams via readStreamableValue

  6. Dynamic UI Updates
    Results appear in real-time with rich, typed AgentPreview components

  7. Autonomous Mode (Optional)
    Auto-launches next evolutionary step in a continuous loop

🏗️ Architecture

flowchart TD
    A["User Input<br/>Text, Logs, Screenshots, Voice"] --> B["/api/forgeflow"]

    B --> C1["runAnalyzer()"]
    B --> C2["runUIArchitect()"]
    B --> C3["runBackendEng()"]
    B --> C4["runJudge()"]
    B --> C5["runScriptwriter()"]

    C1 --> D["MCP Server (/api/mcp)"]
    C2 --> D
    C3 --> D
    C4 --> D
    C5 --> D

    D --> E["parse_log"]
    D --> F["analyze_error"]
    D --> G["generate_fix"]
    D --> H["suggest_v0_prompt"]
Loading

📊 Project Structure

forgeflow/
├── app/
│   ├── layout.tsx              # Root layout with fonts & metadata
│   ├── page.tsx                # Animated landing page
│   ├── dashboard/              # Main ForgeFlow application
│   ├── analytics/              # Session analytics dashboard
│   ├── profile/                # User profile & settings
│   ├── settings/               # Configuration & API keys
│   ├── globals.css             # Global styles & animations
│   └── api/
│       ├── forgeflow/
│       │   └── route.ts        # Fleet orchestration (5 parallel agents)
│       └── mcp/
│           └── route.ts        # MCP server (4 custom tools)
│
├── components/
│   ├── AgentCard.tsx           # Animated agent status cards
│   └── BlueprintPanel.tsx      # Tabbed results viewer
│
├── lib/
│   ├── schemas.ts              # Zod schemas for type safety
│   └── prompts.ts              # System prompts for agents
│
├── package.json
├── next.config.js
├── tailwind.config.ts          # Custom theme configuration
└── tsconfig.json

🧪 Example Use Cases

Debugging Production Issues

Input:

Redis connection refused at port 6379
Error: ECONNREFUSED 127.0.0.1:6379

Output:

  • Root cause analysis with severity score
  • Monitoring dashboard (v0 prompt)
  • Retry logic implementation
  • Custom MCP troubleshooting tool
  • Quality scores from Judge agent
  • 60-second demo script

Building AI Agents

Input:

Build a commit message AI agent that generates semantic commits from git diffs

Output:

  • Complete system architecture blueprint
  • VS Code-style UI design (v0 prompts)
  • Git diff parser MCP tools
  • Agent pipeline configuration
  • Quality evaluation scores
  • Demo walkthrough script

🎯 Production Features

Recently Shipped (v0.1.0)

  1. 📊 MCP Observability & Tracing
    Live waterfall TracePanel tracking tool invocation latency and status

  2. 💬 Slack ChatSDK Integration
    Native /forgeflow slash-command integration with rich insights

  3. 💻 WebContainer Sandbox
    In-browser Node.js execution for safe MCP tool testing

  4. 🏥 Agentic Self-Testing
    HealthScore validation for JSON completeness, security, and code quality

  5. 🧠 Smart Context AI
    Real-time tech stack detection with auto-suggested input modes

  6. 🔗 One-Click GitHub Export
    Auto-generates complete repositories with MCP code, README, and architecture

  7. 📈 Session Analytics
    Full history and error trending powered by Vercel KV

🛡️ Security

ForgeFlow implements enterprise-grade security measures:

App-Level Protection

Security Headers (via next.config.mjs):

  • HSTS: Enforces SSL/TLS for 2 years including subdomains
  • CSP: Prevents XSS by restricting script/style sources
  • X-Frame-Options: Prevents clickjacking (SAMEORIGIN)
  • X-Content-Type-Options: Prevents MIME-type sniffing
  • Referrer-Policy: Protects user privacy

API Key Privacy

  • Local-First Storage: API keys stored in browser localStorage
  • Encrypted Transit: TLS 1.3 encryption for all AI communications
  • Visual Alerts: Blinking notifications for insecure configurations
  • Never Logged: Keys never touch our servers except during active requests

Execution Isolation

  • MCP Sandboxing: Restricted environment prevents unauthorized access
  • WebContainer Isolation: Browser-based micro-VM completely isolated from host

📋 Roadmap

  • ⚡ Parallel agent execution with streaming
  • 🧠 Multimodal input (text, images, logs)
  • 🔌 Custom MCP server integration
  • 🔗 GitHub repository export
  • 👀 Live preview with WebContainers
  • 🤖 Agent self-validation
  • 📊 Session analytics
  • 🔄 Self-evolving agent prompts
  • 🎤 Voice input support
  • 🌐 Multi-language support
  • 🔐 Team collaboration features

🤝 Contributing

Contributions are welcome! Please see CONTRIBUTING.md for guidelines.

Development Setup

  1. Fork the repository
  2. Create a feature branch (git checkout -b feature/amazing-feature)
  3. Commit your changes (git commit -m 'Add amazing feature')
  4. Push to the branch (git push origin feature/amazing-feature)
  5. Open a Pull Request

📝 License

This project is licensed under the MIT License - see the LICENSE file for details.

🏆 Built For

v0 → Agent Hackathon

Transform chaos into clarity.
Turn ideas into shippable AI agents.

Made with ❤️ by the ForgeFlow Team

Live Demo • Report Bug • Request Feature

About

ForgeFlow is an autonomous multi-agent AI engineering fleet built with Next.js and Vercel AI SDK. It deploys 5 parallel AI specialists to instantly debug production logs, architect full-stack blueprints, and generate production-ready UI prompts.

forgeflow-multi-agentic-ai-system.vercel.app

Topics

typescript ai v0 ai-agents tailwindcss vercel-deployment hacakathon vercel-ai-sdk nextjs15 claude-3-5-sonnet muilt-agents vercel-ai-gateway

Resources

Readme

License

MIT license
Activity

Stars

1 star

Watchers

0 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

 
 
 

Contributors

Languages